Programming is generally a good way to train structured thinking and abstraction skills, what is generally good for any scientist. A practical need of programming to solve every-day problems in the lab is rare, to my opinion. If problems requiring programming skills are sporadic or only a minor part of your work, it may be more efficient to co-operate with people who know programming (usually bioinformaticians). If you head for in-depth analysis of your NGS data by yourself, then you actually must learn programming (R or Python). That takes time, and like speaking a language your skills will develop over years of continuous practice.

As a oceanographer , I used R in almost data analyst and visualization purpose.

R is flexible and extensible . It has so many packages . It can be used efficiently in bioinformatics (I saw some of R on bioinformatics) .

Programming is generally a good way to train structured thinking and abstraction skills, what is generally good for any scientist. A practical need of programming to solve every-day problems in the lab is rare, to my opinion. If problems requiring programming skills are sporadic or only a minor part of your work, it may be more efficient to co-operate with people who know programming (usually bioinformaticians). If you head for in-depth analysis of your NGS data by yourself, then you actually must learn programming (R or Python). That takes time, and like speaking a language your skills will develop over years of continuous practice.

R programming is very important and can be used easily and you can learn it as well and easily

I have less than 10 years to retirement so I think I'll pass and accept that life is more than statistics. I'll have to manage the best I can with my old software.

Best

JK

Useful.

I feel it is a must for all researchers as it is free.

I agree with Jochen Wilhelm. Unless you are working in a data center, using programming to solve every-day problems in a lab will be rare. If you really want to learn programming I think python is the best start point, since there is a lot of thing in R that is more suitable for statistical analysis.

HI Sinan

I'm just accustomed to Mplus and SPSS, "old", familiar software.

Best

Jk

I would add that, beside the need for clear and logical mind always suitable for a scientist, you can have to program for two main reasons:

- automatize some data raw results pre-treatment, typically reformat results file in another format more suitable for analysis or compute some secondary data from the raw data

- do some "unusual " things with your data, unusual meaning here not available in your usual analysis software.

In the first case, python can be a good choice, as other languages.

In the second case, and partly in the first case for some kinds of pre-treatment, typically for NGS data, programming is needed, but is not sufficient: you must also understand what the method is doing, that is at least partly the underlying mathemical/physical/chemical models (for data normalization, for the most simple example, but can be also for the use of advanced statistical methods). If it's for a punctual experiment, you may have more efficient results by contacting specialists. Knowing how to program R to load a package and call its functions correctly does not mean the results they give has a meaning...

Yes, Emmanuel, these are relevant applications of programming. But these are just no examples demonstrating the need to learn programming yourself. Automatation is usually a job you do once, and then the system runs. It's finally cheaper and faster to hire a programmer to set up the automatation. If you have one of the jobs where you frequently have to automat different processes, then you should better be able to programm, for sure! - to my experience that's not often the case for life-scientists (those people typically interested in NGS). Doing "unusual" things also implies (at least to me) that these are things that are not done very often. Here I would also consider a cooperation with a programmer (or a scientist with sufficient programming skills) more efficient.

In my opinion, every scientist should have at least basic understanding of R programming. Previous responses are correct that automation, in theory, precludes the need for ongoing coding efforts. However, analytic tools and benchmarks are constantly evolving. In addition, we have entered an era of "Big Data" in science, where the generation of 'omics analyses is increasingly common and even expected. Very soon, however, the integration and reanalysis of existing (i.e. published) datasets will become more important, since NIH-funded data are publicly available. In fact, the NIH Big Data to Knowledge (BD2K) emphasizes this need to incorporate public data repositories into ongoing studies.

As Jochen states, collaborating with a bioinformatician is definitely advised, but designing your experiments and understanding the results both require a basic understanding of programming.

Someone also said python is a great starting place, but I personally disagree. Although Python is fantastic for assembling pipelines, R is WAY better place to start for basic scientists, mainly because it is a statistical programming language with the ability to wrangle and visualize data. The RStudio (https://www.rstudio.com) interface is probably easiest for most scientists wanting to learn programming (allows you to organize your work and see what your scripts are doing).

I think , it depends on your work. R is opensource and widely used for statistical analysis and visualization. but excel is also btter opiton . R needs coding experiences and basic understanding of structure of programming. R have numerous pkgs , they are the real things for research .

@ Hafez Ahmad: basic Excel has very few statistical capabilities, and (unless corrected in the last version) is known to use very poor algorithms to compute such basic things as variances, leading to rounding issues (that can even give negative variances in some cases!). So I would never say that Excel is « better » option for anything else than, eventually, creating the data table.

Of course, by using adequate extensions, Excel can become very good for statistics, but that's another debate.

to add:

Excel had many problems doing incorrect analyses in versions older then 2010. In the 2010 version, many errors have been corrected. But still, important and quite basic issues are not resolved in the current version, which are:

• inconsistent interpretation of missing values,
• wrong probability plots (Y scale in the normal-QQ plot is linear)
• wrong residual plots (using the data rather than the predicted values)
• probability distributions are evaluated with suboptimal algorithms giving only a limited precision (that is, p-values can be wrong after the 2nd decimal digit)
• regression functions still return plain wrong results in the presence of multicollinearity
• Excel uses its SOLVER to get approximate maximum-likelihood estimates. The SOLVER is far from working robustly and too often returns only a local optimum

No need to say that Excel does rather facilitate rather than obviate mixing data and processing/analysis.

Hi,

R is very robust with thousands of packages deals with statistical, computing, visualization expression studies and NGS packages as well.

First of all, if you find it difficult to understand by reading the manuals, you may see some video lectures on R programming, which will give you a practical experience as well as understanding.

Hi Heba Huessin ,

Mastering R will help you to solve many problems in your research.

When you deal with massive NGS data, you will find yourself in need of automation tools.

When I started analysing my NGS data, my first option was to learn python, but without programming language background, it was difficult for me, then I came across R.

I found R is comparatively easier than python.

Many packages are available in R to deal with NGS data.

R and its packages are free, and have a large community of users who are ready to help you if you got struck with scripts.

In case of NGS (if you dealing with microbial data) you can find R as one stop solution for all your needs.

Mostly R is used for statistical analysis. But some packages allow you to do more. For eg:the DADA2 package can be used for entire Microbial 16s marker NGS data analysis from merging paired reads to construct an OTU table (in dada2 instead of OTUs a more accurate ASVs are used.)

R gives you ability to create data visualisations with ease.

ggplot is a widely used package for creating different types of plots, which you can try as beginner.

Hey Heba Huessin

R Programming language is one of the most demanded programming languages for researchers as it has a solution to most of their problems.

Here are a few more reasons-

I believe that in the current scientific era a basic knowledge of either R or python is a necessary evil for all scientists. Although I agree with the previous answers that emphasised hiring a bioinformatician or biostatistician, it is not always the best option. Sometimes a deeper knowledge of a specific scientific domain is need in order to accurately analyse a particular dataset. Some Expert programmers may luck that knowledge and hence lead to erroneous analysis and interpretation of results.

Precious Kwadzo Pomary : hiring a bioinformatician or a biostatistician does not mean to give him data and a question and let him alone with that... Only a close interaction can give meaningful results, because of the knowledge of the scientific domain needed (for the statistician/informatician) *and* the knowledge of the technics (either stats or programmation, for the scientist).

But I agree that a basic knowledge of a programming language is a huge advantage for a scientist. R, python or another...

I agree, just "hiring" a bioinformatician or biostatistician usuaylly is not very fruitful. There must be a tight cooperation between biologist (or physician), bioinformatician and biostatistician. All should have enough knowledge on all three topics to find a common language and to really understand each other. That's all hard work, still - and in addition to [some] subject matter knowledge in all fields communication skills are required.

And for many "simpler" problems, the skills needed to properly discuss the problem with a bioinformatician and a biostatistician turns out to be sufficient to solve the problem alone (a bioinformation can then still be helpful to develop an scalable and automatable solution, a biostatistician can the still be helpful to generalize the solution or to propose more efficient or more robust models).

A lot of this discussion seems R versus Excel. I like R, but I would assume the alternatives would be things like Python, Julia, MATLAB, SAS, STATA, etc. If the questioner wants to learn a stats program/language/environment, it might be worth thinking which is most appropriate.